A Hamilton-Jacobi approach of sensitivity of ODE flows and switching points in optimal control problems

TL;DR

This paper develops a Hamilton-Jacobi-based method to compute sensitivities of ODE flows and switching points in optimal control problems, improving the analysis of bang-bang and singular solutions.

Contribution

It introduces a new formula for derivatives of ODE flows with respect to initial conditions, tailored for Hamilton-Jacobi-Bellman equation applications.

Findings

Derived a practical formula for ODE flow sensitivities

Applied results to analyze switching times and reachable sets

Enhanced verification of optimality in control problems

Abstract

In optimal control problems of control-affine systems, whose solutions are bang-bang or singular type, verification of optimality using the Hamilton-Jacobi-Bellman (HJB) equation involves the computation of partial derivatives of switching times and switching states with respect to initial conditions (time and state). In this paper, we establish a formula for the partial derivatives of ordinary differential equations (ODE) flows with respect to initial conditions, which is more suitable for using in HJB equation than such provided by the classical theory of ODE. We apply the obtained results to the sensitivity analysis of hitting time and state of a reachable set, that in an optimal control problem can represent a switching locus.